A Bibliometric Analysis of Research on Application of AI in Wastewater Treatment, 1987-2024

doi:10.21203/rs.3.rs-8000973/v1

AI is now an essential driver of innovation and improvement of wastewater treatment systems due to its ability to use data-driven management for more sustainable operations, reduced energy use, and more effective monitoring of contaminant occurrence. In the last few decades, AI tools have shifted from a theoretical basis to principles of applied research in the world of environmental engineering. This study provides a bibliometric analysis of 2,096 journal articles published between 1987 and 2024 that were obtained from Scopus and Web of Science databases. The analysis used Bibliometrix R package versions examining the intellectual perspective, geographic perspective, and temporal of worldwide AI applications studies focusing on wastewater treatment. The analysis showed exponential growth after 2016 with a steady annual increase of 17.14% confirming rapid growth in this interdisciplinary topic area. China provided the most annual publication output; whereas, Spain provided the most cited articles indicating a varied sense of national priorities in achieving simultaneous publications while achieving impactful research. Saudi Arabian and Chinese universities displayed the most institutional membership. King Fahd University of Petroleum and minerals was the most productive institution in the overall publication dataset obtained. The thematic shift illustrated a move from more narrow model-specific research focused on “artificial neural networks” towards broader models based on more generalized conceptualizations of “machine learning” and “deep learning.” This shift illustrated an overall awareness of the expanding role of artificial intelligence in optimization and sustainability to research processes regarding waste treatment methods. As a whole, this research provides a data-driven perspective of how AI-related research has developed overtime as well as an overview of global collaboration networks and potential research directions for the future to promote a circular economy and sustainable water management.

Conservation Biology

Agricultural Engineering

Renewable Resources

Environmental Engineering

Developmental Biology

Artificial Intelligence

Machine Learning

Wastewater Treatment

Bibliometric Analysis

Deep Learning

Scientometrics

In recent decades, increasing industrial activity, urbanization, and intensive agriculture have put strong pressure on global freshwater supply. Consequently, wastewater treatment is now an important component of sustainable water management and environmental protection; pollutants continuous entry into bodies of water, new contaminants have emerged, and water demand and supply is increasingly imbalanced, all contributing to shift wastewater treatment from an unimpressive technical task to a neighborhood sustainability concern that encompasses multiple dimensions of the environment [1, 2]. Conventional biological, chemical, and physical techniques for wastewater treatment are well known and have been applied for many years. However, these methods often depend on stable conditions, the operator’s experience, and high energy consumption, which limit their flexibility and performance [3]. The challenge becomes even greater in modern wastewater systems, where influent quality and pollutant loads vary constantly and must still comply with strict environmental standards [4]. For these reasons, researchers and engineers are now focusing on innovative technologies capable of improving treatment efficiency, reducing operational uncertainty, and supporting sustainable plant management.

Among the emerging advanced technologies, Artificial Intelligence (AI) is a powerful solution to these challenges. AI tools empower researchers and engineers to discover patterns, make predictions about the future, and optimize operations with large datasets [4, 5]. While mechanistic models rely on explicit equations in mathematical structures, AI models learn directly from data, and are particularly adept at incorporating complex nonlinearities found in biological and physicochemical processes found in wastewater [6]. The adaptability and learning directly from data enables automation to critical tasks such as effluent quality prediction, fault detection, and process control [7, 8]. For instance, deep learning architectures have facility for excellent anomaly detection in sewer systems and detect microbial community composition from WTP metagenomic datasets [9, 10, 11]. This movement toward predictive intelligence will remold wastewater treatment plants WWTP practices toward intelligent systems that self-monitor human systems, and real time optimize, such as through predictive maintenance.

Beyond its technical applications, Artificial Intelligence (AI) is now recognized as a key enabler connecting wastewater management with broader goals of sustainability and the circular economy. This new perspective views wastewater not as a pollutant but as a valuable source of water, energy, and nutrients, aligning with global trends toward resource recovery and reuse [12, 13]. AI plays an important role in these transitions by supporting multi-objective optimization, life-cycle analysis, and adaptive control of treatment systems that operate under variable and often harsh environmental conditions [14, 15]. For example, AI models have been used to estimate biogas production, improve the operation of membrane bioreactor systems, and guide nutrient recovery from waste streams using both deterministic and probabilistic models [16, 17]. Also, the integration of AI and big data infrastructure provides additional support to decentralized and resilient wastewater systems where the intelligence for decision-making is distributed and rapidly iterated on real time [18, 19]. Altogether, these developments contribute to “smart wastewater management,” where digital tools complement human expertise to achieve efficient, reliable, and sustainable performance.

Recent studies emphasize that Artificial Intelligence (AI) represents more than a simple technological upgrade; it marks a major shift in how wastewater systems are understood and managed. Data-driven and self-learning algorithms have replaced traditional fixed-parameter systems, allowing for stronger process optimization and improved environmental protection [20]. Through AI, wastewater treatment plants (WWTPs) are evolving toward predictive and adaptive operations, capable of anticipating sudden variations in influent characteristics and adjusting performance in real time to minimize environmental impacts [21]. For example, Manhou et al. (2024,2025) [22, 23, 24], demonstrated that saline irrigation influences crop performance and soil quality, highlighting how data-driven approaches can optimize irrigation management and wastewater reuse. These examples show a growing intersection between environmental engineering, agricultural science, and artificial intelligence, where smart algorithms improve resource use efficiency and enhance resilience to environmental stresses. In addition, the development of explainable AI methods helps ensure that model predictions are understandable for plant operators and regulators two groups often hesitant to adopt opaque technologies. Consequently, AI is increasingly seen not as a “black box,” but as a valuable decision-support tool that complements and strengthens existing engineering practices [25].

The expansion of AI in wastewater treatment has been fueled by major advances in computing power, cloud-based data storage, and sensor technologies [5]. Modern monitoring devices now produce continuous datasets describing flow rates, effluent concentrations, and operational conditions, providing rich material for training and validating AI models [26]. Early applications in the 1980s and 1990s relied mainly on artificial neural networks (ANNs) and genetic algorithms for process optimization and energy management [27, 28]. In the 2000s, hybrid AI models combining data-driven and statistical or mechanistic approaches appeared [6]. After 2016, the field experienced a major acceleration due to open-access data, machine learning frameworks like TensorFlow and PyTorch, and global interest in climate resilience [17, 11]. This rapid growth reflects the increasing maturity and acceptance of AI as a fundamental methodology in water and environmental research.

While all this growth in research has clearly been beneficial, it has caused challenges surrounding knowledge fragmentation and overlapping disciplines. Numerous studies occur in the literature regarding different subtopics ranging from modeling pollutant adsorption and forecasting energy use, to developing hybrid bioelectrochemical systems [2, 3]. The vast diversity of methods and contexts ( experimental versus simulation studies, rural versus urban contexts) makes it increasingly difficult for researchers as well as the broader community to identify common research trends, and ultimately, researchers have begun to rely on bibliometrics as a method to systematically map and assess the field. Bibliometrics provides a quantifiable signal for research productivity, thematic evolution, and collaborative networks across research articles, thus illuminating the underlying intellectual structure of a rapidly growing research area [29]. By means of citation analysis, co-authorship networks, and keyword mapping, it is feasible to examine how researchers conceptual relationships among AI methodologies with sewered and unsewered wastewater methodologies, and to identify leading institutions, countries, and authors in the area of AI innovations.

The geographical distribution of research on AI applications in wastewater management indicates different national priorities and approaches. Bibliometric data suggest that China is the leader in terms of research volume, while fewer papers produced by countries such as Spain and Iran produce greater citation impact, suggesting an imbalance between publications and uptake [30, 15]. Similarly, Saudi Arabia and Australia institutions have strong international networks of collaboration to share knowledge globally [14]. These findings suggest that the development of AI research in wastewater treatment varies across regions and depends largely on each country’s scientific infrastructure, policy direction, and industrial priorities.

By bringing these aspects together, this research provides a well-rounded picture of how artificial intelligence has been integrated into scientific, technological, and policy contexts of wastewater management. It locates artificial intelligence not simply as a tool of computing, but as an integral aspect of environmental innovation, which has emerged as a new path for the enabling of optimized, automated and sustainable wastewater systems throughout the world. In the end, this bibliometric analysis aims to provide a robust foundation for future research that will further both academic research and applied science for intelligent, adaptive, and efficient water management systems for a circular, sustainable future.

2.1 Data Source and Search Strategy

This study relied on two multidisciplinary databases Scopus and the Web of Science (WoS) Core Collection to ensure broad and comprehensive coverage of relevant literature. The WoS search included all major indexes (SCI-Expanded, SSCI, A&HCI, CPCI-S, CPCI-SSH, BKCI-S, BKCI-SSH, and ESCI).

The search was executed on February 18, 2024, targeting studies that explore the intersection between artificial intelligence and wastewater treatment. The query applied to both databases was formulated as follows:

("artificial intelligence" OR "machine learning" OR "deep learning") AND ("wastewater" OR "sewage" OR "reclaimed water" OR "recycled water" OR "water reuse" OR "water recycling").

In Scopus, searches were limited to the Title, Abstract, and Keywords fields, and filtered for English-language journal articles published up to 2024. In the Web of Science Core Collection, the query was applied to the Topic field which includes the title, abstract, and author keywords under the same filters for document type and language.

2.2 Data Collection, Curation, and Selection

The first search produced 1,936 records in Scopus and 1,559 records in the WoS Core Collection. Metadata (titles, authors, author affiliations, abstracts, keywords, year of publication, journal, citation counts) for all records were exported from both databases.

These datasets were then imported into the Bibliometrix R package (version 4.3.3) within the RStudio environment (version 2024.12.1) for processing and analysis. During the import process, a slight refinement occurred, resulting in 1,917 records from the Scopus dataset and 1,484 records from the WoS dataset being successfully loaded into Bibliometrix.

Data integration and cleaning were performed using the Bibliometrix package in RStudio (v2024.12.1). The mergeDbSources command was employed to merge both datasets, followed by a duplicate detection step using duplicatedMatching to eliminate overlapping entries. After removing 1,266 duplicates, a total of 2,135 unique documents were retained.

Publication dates were verified to ensure consistency with the study period. Articles indicating 2025 as the publication year (typically “in press” or early online versions) were excluded, leaving only studies published through 2024. The final data collection and cleaning process are summarized in Fig. 1, which outlines each step of screening and refinement.

2.3 Bibliometric Analysis and Visualization Tools

All analyses were conducted using the Bibliometrix R package (version 4.3.3). This software enabled the computation of key bibliometric indicators and the generation of visual data representations.

The analysis consisted of two main components:

Performance Analysis

Evaluating publication trends over time (Annual Scientific Production), identifying the most productive and impactful authors (Most Relevant Authors), institutions, and countries, and determining the core publication outlets (Most Relevant Sources/Journals).

Science Mapping

Analyzing the conceptual structure of the research field through keyword co-occurrence analysis, and examining collaborative patterns through co-authorship network analysis.

While Bibliometrix was used for generating the core data and most visualizations, the geographical distribution of publications by country was visualized using the online tool Mapchart (mapchart.net). The underlying country production data for this map was extracted directly from the analysis performed in Bibliometrix. The specific results and visualizations generated are presented in the Results section.

The bibliometric analysis was conducted on a final dataset comprising 2,096 unique journal articles focused on the intersection of Artificial Intelligence (AI) and wastewater treatment, published up to the end of 2024. The data was retrieved from Scopus and Web of Science Core Collection databases, following the curation process detailed in the Methodology section and illustrated in Fig. 1. This section presents the key findings derived from this dataset.

A summary of the main bibliometric characteristics of the dataset is presented in Table 1. The analyzed literature spans from 1987 to 2024, indicating the earliest identified work in this specific cross-section of AI and wastewater treatment. These 2,096 articles were published across 616 different scientific sources (journals). The field has demonstrated significant dynamism, evidenced by an annual growth rate of 17.14% over the period where publications appeared consistently.

The research involves a substantial community, with 7,411 authors contributing to the publications. Collaboration appears to be a dominant feature of this research area. This is highlighted by the low number of single-authored documents (only 47) and a high average number of co-authors per document (5.64). Furthermore, international collaboration is prominent, with 25.1% of the documents involving co-authorship between researchers from different countries.

The dataset encompasses 5,970 unique author keywords (DE), reflecting the thematic diversity within the field. The average age of the documents in the dataset is relatively low at 4.01 years (calculated relative to the search date), suggesting that the literature is, on average, quite recent. Finally, the publications in this dataset have received an average of 17.63 citations per document, indicating a notable level of scientific impact and visibility within the academic community.

Table 1

Main Information about the Bibliometric Dataset
Metric	Value
Timespan	1987–2024
Sources	616
Documents	2096
Annual Growth Rate	17.14%
Authors	7411
Authors of Single-Authored	47
International Co-Authorship	25.1%
Co-Authors per Document	5.64
Author’s Keywords (DE)	5970
Document Average Age	4.01
Average Citations per Doc	17.63

3.1 Annual Scientific Production and Citation Trends

To understand the temporal evolution and impact of research in this field, the annual scientific production and corresponding citation counts were analyzed. Figure 2 presents the distribution of publications (blue bars, left axis) and the total citations received by those publications per year (red line, right axis) from 1987 to 2024.

The analysis reveals a distinct growth pattern. The field emerged sparsely, with the first publications appearing in 1987 (n = 2). Activity remained very low and sporadic for nearly two decades, typically fewer than 10 articles published annually until 2008. A phase of slow but steady growth began around 2008, with publication numbers consistently reaching double digits.

There was a notable acceleration in publications beginning in around 2016 (n = 33). The acceleration was particularly strong, and near-exponential, from 2019 onward. Publications nearly doubled between 2020 (n = 98) and 2021 (n = 209). The volume of publication crossed the 200 mark for the first time in 2021, then grew to nearly 300 (n = 299) in 2022 and surpassed 400 (n = 416) in 2023. 2024 records the steepest increase in the total number of publications recorded in this dataset, totaling 697. This acceleration noticeable in recent years indicates the growing interest and activity using AI methods for wastewater treatment related challenges.

The trend of annual citations generally mirrors the publication trend, albeit with an expected time lag inherent to the citation process. Citations remained low during the early years but started to climb significantly after 2017, corresponding to the increased publication volume and growing recognition of the field. The peak in annual citations occurred for papers published in 2021 (6,399 citations). The subsequent decrease in total citations for publications from 2022, 2023, and 2024 is a typical bibliometric artifact, as these more recent articles have had less time to accumulate citations compared to older ones. Nonetheless, the substantial number of citations accrued even by these recent papers highlights the ongoing impact and relevance of the research being published.

3.2 Most Relevant Sources

The dissemination channels for research on AI in wastewater treatment were identified by analyzing the journals that published the articles in the dataset. Table 2 lists the top 10 most productive journals based on the number of publications (TP), along with metrics reflecting their impact both within this specific dataset (TC, TC/TP, h-index) and more broadly (Impact Factor 2023, CiteScore 2023) and their primary subject categories.

Table 2

Top 10 Most Relevant Sources Publishing Research on AI in Wastewater Treatment (1987–2024)
Rank	Journal	TP¹	TC²	(TC/TP)³	h-index (Dataset)⁴	IF 2023⁵	CS 2023⁶	Categories/positions (Exemplary)⁷	ISSN numbers	Initial year
1	SCIENCE OF THE TOTAL ENVIRONMENT	95	1743	18.35	23	8.2	17.6	Env Sci, Env Eng (Q1/9/197); Env Sci, Pollut (Q1/9/167); Env Sci, Waste Mgt (Q1/10/134)	0048-9697	1972
2	JOURNAL OF ENVIRONMENTAL MANAGEMENT	85	1530	18	22	8	13.7	Env Sci, Mgt/Policy (Q1/17/399); Env Sci, Env Eng (Q1/13/197); Env Sci, Waste Mgt (Q1/15/134)	0301–4797	1973
3	JOURNAL OF WATER PROCESS ENGINEERING	70	645	9.21	13	6.3	10.7	Chem Eng, Process Tech (Q1/12/73); Env Sci, Waste Mgt (Q1/24/134); Biotech (Q1/42/311)	2214–7144	2014
4	WATER RESEARCH	65	1455	22.38	23	11.5	20.8	Env Sci, Water Sci/Tech (Q1/2/261); Env Sci, Env Eng (Q1/5/197); Env Sci, Waste Mgt (Q1/6/134)	0043-1354	1967
5	CHEMOSPHERE	63	1451	23.03	23	8.1	15.8	Env Sci, Env Chem (Q1/14/147); Env Sci, Env Eng (Q1/11/197); Env Sci, Pollut (Q1/12/167)	0045-6535	1972
6	WATER SCIENCE AND TECHNOLOGY	59	795	13.47	15	2.5	4.9	Env Sci, Water Sci/Tech (Q2/86/261); Env Sci, Env Eng (Q2/77/197)	0273–1223	1969
7	JOURNAL OF CLEANER PRODUCTION	49	1683	34.35	25	9.8	20.4	Env Sci, Gen Env Sci (Q1/5/233); Energy, Renew/Sustain/Env (Q1/17/270); Eng, Indust/Manuf (Q1/10/384)	0959–6526	1993
8	BIORESOURCE TECHNOLOGY	43	812	18.88	15	9.7	20.8	Env Sci, Env Eng (Q1/6/197); Env Sci, Waste Mgt (Q1/7/134); Energy, Renew/Sustain/Env (Q1/16/270)	0960–8524	1991
9	WATER	38	514	13.53	12	3	5.8	Env Sci, Water Sci/Tech (Q1/63/261); Agri/Bio Sci, Aquatic Sci (Q1/40/247)	2073–4441	2009
10	ENVIRONMENTAL SCIENCE AND POLLUTION RESEARCH	31	411	13.26	13	N/A	8.7	Env Sci, Pollut (Q1/31/167); Env Sci, Env Chem (Q1/33/147); Env Sci, Health/Tox/Mutagen (Q1/25/148)	0944–1344	1994
1 TP: Total Publications in dataset
2 TC: Total Citations from dataset
3 TC/TP: Average Citations per Publication

4 h-index (Dataset): h-index based on TP & TC within the dataset

5 Journal Impact Factor for the year 2023. N/A indicates the value was not available from the source checked

6 CiteScore for the year 2023

7 Categories/Positions (Exemplary): Shows selected subject categories relevant to the study's topic, indicating journal quartile, rank within category, and total journals in category (Format: Qx/Rank/Total). "(Exemplary)" signifies that this is a selection, not necessarily an exhaustive list of all categories for the journal. Category Abbreviations: Agri/Bio Sci = Agricultural and Biological Sciences; Biotech = Biotechnology; Chem Eng = Chemical Engineering; Env Chem = Environmental Chemistry; Env Eng = Environmental Engineering; Env Sci = Environmental Science; Gen Env Sci = General Environmental Science; Geog/Plan/Dev = Geography, Planning and Development; Health/Tox/Mutagen = Health, Toxicology and Mutagenesis; Indust/Manuf = Industrial and Manufacturing Engineering; Mgt/Policy = Management, Monitoring, Policy and Law; Pollut = Pollution; Process Tech = Process Chemistry and Technology; Renew/Sustain/Env = Renewable Energy, Sustainability and the Environment; Waste Mgt = Waste Management and Disposal; Water Sci/Tech = Water Science and Technology.

Science of the Total Environment emerged as the leading journal, publishing the highest number of articles (TP = 95) in this field. It is closely followed by the Journal of Environmental Management (TP = 85) and the Journal of Water Process Engineering (TP = 70). These three journals collectively account for a substantial portion (250 articles, approx. 11.9%) of the total publications, indicating their central role in disseminating research at the intersection of AI and wastewater treatment.

Several other journals also serve as key outlets, including Water Research (TP = 65) and Chemosphere (TP = 63). While ranking slightly lower in publication volume within the top 10, some journals demonstrate high impact. Notably, the Journal of Cleaner Production exhibits the highest average citations per paper (TC/TP = 34.35) and the highest dataset-specific h-index (h = 25), suggesting its published articles in this area are particularly influential, despite having fewer publications (TP = 49) than the top-ranked journals. Water Research (TC/TP = 22.38, h = 23) and Chemosphere (TC/TP = 23.03, h = 23) also show high citation impact per paper specific to this topic. The high overall IF (11.5) and CS (20.8) for Water Research, and the high CS for Bioresource Technology (20.8) and Journal of Cleaner Production (20.4), further underscore their prominence.

An analysis of the journal categories indicates that most published research is featured in journals addressing Environmental Science, Environmental Engineering, Water Science and Technology, Waste Management, and Pollution. The journals Science of the Total Environment, Journal of Environmental Management, Water Research, and Chemosphere fall under core outlets for environmental and water research. The Journal of Water Process Engineering signifies the role of chemical and process engineering contexts, while the Journal of Cleaner Production and Bioresource Technology expand the context to sustainability and resource management and bioengineering perspectives. Water Science and Technology (primarily Q2) clarifies that relevant research is also available in reputable journals with specific, respected impact factors, despite being lower than the top-tier Q1 journals at the top of the list. Most of the top-10 journals are ranked as Q1 journals in their respectivemost relevant category, indicating that research on AI for wastewater treatment is frequently published in high-impact, reputable outlets related specifically to environmental science and engineering.

3.3 Geographic Distribution and Performance of Countries

To understand the global landscape of research on AI in wastewater treatment, the geographic distribution of publications and the performance of contributing countries were analyzed. The analysis considered total publication output, citation impact, and patterns of national and international collaboration.

3.3.1 Introduction & Overall Productivity

The overall geographic distribution of publications is visualized in Fig. 3, which maps the total number of publications (TP) attributed to authors' affiliations across different countries. Countries are colored based on publication volume according to the legend.

The map reveals a wide international interest in the field, with contributions originating from numerous countries across all continents. However, research productivity is highly concentrated in certain regions. Asia emerges as the most prolific continent, largely driven by China, which stands out with a significantly higher publication volume (TP = 2,072) than any other country. North America, represented primarily by the USA (TP = 686), is another major center of research. Significant contributions also originate from other Asian countries like India (TP = 530), Saudi Arabia (TP = 400), Iran (TP = 290), and South Korea (TP = 253), as well as several European nations like Spain (TP = 191) and the UK (TP = 180), and Australia (TP = 154) in Oceania. Many countries in Europe, South America, and Africa show moderate to low levels of activity, indicating varying degrees of engagement with this research topic globally.

For a deeper quantitative analysis of country performance, Table 3 displays the top 10 most productive countries ranked by Total Publications (TP). It also presents the citation impact variables (Total Citations - TC; Average Citations per Publication - TC/TP) and collaboration (Corresponding Author papers, Single Country Publications - SCP; Multiple Country Publications - MCP, and the MCP rate as a percentage). These aspects will be analyzed further in subsequent sections.

Table 3

Performance and Collaboration Metrics for the Top 10 Countries Contributing to AI in Wastewater Treatment Research (1987–2024)
Rank	Country	TP	TC	TC/TP	Corr. Auth.⁸	SCP⁹	MCP¹⁰	MCP Rate (%)¹¹
1	CHINA	2072	9340	4.51	601	488	113	18.8%
2	USA	686	4485	6.54	174	142	32	18.4%
3	INDIA	530	1818	3.43	158	132	26	16.5%
4	SAUDI ARABIA	400	933	2.33	70	30	40	57.1%
5	IRAN	290	1910	6.59	97	65	32	33.0%
6	SOUTH KOREA	253	1591	6.29	74	53	21	28.4%
7	SPAIN	191	1999	10.47	81	69	12	14.8%
8	UK	180	695	3.86	44	30	14	31.8%
9	CANADA	176	719	4.09	56	42	14	25.0%
10	AUSTRALIA	154	682	4.43	43	23	20	46.5%

8 Number of publications where the corresponding author is affiliated with the country (measure of research leadership).

9 Single Country Publications (number of Corr. Auth. papers involving authors only from that country).

10 Multiple Country Publications (number of Corr. Auth. papers involving authors from at least one other country).

11 Percentage of corresponding author papers involving international collaboration ((MCP / Corr. Auth.) * 100).

As indicated on the map and detailed in Table 3, China is the clear leader in publication volume (TP = 2,072), followed by the USA (TP = 686) and India (TP = 530). These top three countries alone account for a substantial share of the total research output in this domain.

3.3.2 Citation Impact

Examining the citation impact metrics in Table 3 provides further insights into the influence of contributions from different countries. In terms of Total Citations (TC), the ranking largely mirrors the productivity ranking for the very top positions. China leads significantly with 9,340 citations, followed by the USA with 4,485 citations. However, some variations emerge further down the list. Notably, Spain (ranked 7th in TP) achieves the 3rd highest total citations (TC = 1,999), surpassing India, Saudi Arabia, Iran, and South Korea despite having lower publication output. Similarly, Iran (5th in TP) also garnered substantial citations (TC = 1,910), placing it 4th by this metric.

A more nuanced view of citation impact, normalized by publication volume, is offered by the Average Citations per Publication (TC/TP). Here, Spain stands out prominently with the highest average impact (TC/TP = 10.47), suggesting that its publications in this field are highly cited on average. Iran (TC/TP = 6.59), the USA (TC/TP = 6.54), and South Korea (TC/TP = 6.29) also demonstrate strong citation impact relative to their output, achieving notably higher average citations per paper compared to several other leading nations in this list. Conversely, countries like India (TC/TP = 3.43) and Saudi Arabia (TC/TP = 2.33), despite their high productivity (ranked 3rd and 4th in TP respectively), show lower average citation impact per paper compared to other leading nations within this top 10 list. This indicates that while contributing significantly in volume, the average citation influence of their papers in this specific dataset is comparatively lower than that of publications from countries like Spain or Iran. China's TC/TP (4.51) is moderate relative to the other top performers in this group, suggesting its leading position in total citations is primarily driven by its vast publication volume.

3.3.3 Leadership & Collaboration Intensity

Beyond overall productivity and citation impact, the analysis explored patterns of research leadership and international collaboration. Figure 4 presents the distribution of publications based on the country affiliation of the corresponding author, differentiating between Single Country Publications (SCP), involving authors only from the corresponding author's country, and Multiple Country Publications (MCP), which include authors from at least one other country. Table 3 provides the specific counts (Corr. Auth., SCP, MCP) and the calculated MCP Rate (%), indicating the proportion of internationally collaborative papers led by authors from that country.

The ranking based on corresponding authorship (Table 3, Corr. Auth. column) generally aligns with the overall productivity ranking, with China (601 papers), the USA (174 papers), and India (158 papers) leading in the number of publications where they hold the corresponding author role. This suggests these countries are not only high producers but also frequently lead the research projects reported.

Figure 4 and the MCP Rate (%) in Table 3 reveal diverse patterns of international collaboration leadership among the top countries. Some nations demonstrate a high propensity for leading internationally co-authored studies. Saudi Arabia exhibits the highest rate, with 57.1% of its corresponding-authored papers involving international partners (MCP). Australia (46.5%), Iran (33.0%), and the UK (31.8%) also show substantial rates of leading collaborative international publications. This indicates a strong outward-looking collaborative strategy or integration into international networks for researchers in these countries when they are leading projects in this field.

Conversely, several other highly productive countries show a stronger tendency towards domestic leadership or collaboration. Spain has a notably low MCP Rate (14.8%), suggesting that while impactful (as seen from its high TC/TP), its research leadership in this field is predominantly within nationally contained projects. Similarly, major players like China (18.8%), the USA (18.4%), and India (16.5%) have relatively lower MCP rates compared to countries like Saudi Arabia or Australia, indicating that a large proportion of the research led from these nations involves domestic collaborators primarily.

3.3.4 Collaboration Network Structure

To further elucidate the structure of international cooperation, a co-authorship network map based on country collaborations was generated, as shown in Fig. 5. In this network, nodes represent countries, and the links between them indicate co-authorship on publications within the dataset. The size of the nodes corresponds to the number of publications or collaborative links, while the thickness of the links reflects the frequency of collaboration between two countries. Colors are used to delineate clusters of countries that collaborate more frequently with each other than with countries outside their cluster. Clusters identified using the Walktrap algorithm.

The collaboration network visually reveals a distinct core-periphery structure. China appears prominently positioned near the center of the map with the largest node size and numerous connections, suggesting a central role in the network. The USA also occupies a visually central position with a large node size and many links. Saudi Arabia is noticeable for its relatively large node size and strong connections, particularly bridging towards a distinct cluster of countries. Other countries like the UK, Australia, Canada, and India are also visibly well-connected within the main network structure.

The visualization highlights several distinct collaborative clusters based on proximity and link color:

The largest cluster (Red) appears centered around China and the USA, encompassing major contributors from North America (Canada), Europe (UK, Germany, Sweden, Netherlands), Asia (India, Korea, Japan, Vietnam, Singapore), and Africa (South Africa, Nigeria). This visually represents the main global research nexus.
A prominent secondary cluster (Green) visibly connects Saudi Arabia with Malaysia, Egypt, Pakistan, Iraq, UAE, Algeria, Indonesia, and Russia, suggesting strong regional collaborations.
A Blue cluster links European nations like Spain, Portugal, and Austria.
An Orange cluster connects Italy, Poland, and Brazil.
Other smaller groups and more isolated nodes (e.g., Israel, Morocco) indicate countries with fewer observed collaborations within this network visualization.

Visually strong collaboration pathways (indicated by thicker links) appear between China and the USA, China and Saudi Arabia, Saudi Arabia and Malaysia, and within European and North American subgroups. Overall, the visual structure of the network underscores the prominent role of a few key countries, particularly China and the USA, in international collaboration, alongside evidence of distinct regional collaborative groups.

3.3.5 Temporal Dynamics

To understand the historical context leading to the current landscape, the evolution of publication output over time for the five most productive countries is presented in Fig. 6.

Figure 6 highlights distinct development trajectories among the leading nations. The USA was among the earliest and most consistent contributors during the initial decades of research in this field. In contrast, China's involvement began later but experienced near-exponential growth, particularly visible from approximately 2018 onwards. This rapid acceleration resulted in China surpassing all other nations to become the most prolific publisher in recent years. Significant growth in publication output, especially since 2017–2018, is also evident for other leading Asian countries shown, namely India, Iran, and Saudi Arabia, highlighting a recent intensification of research activity in these nations alongside the established players.

3.4 Performance of Research Institutions

To identify the key institutional players and their collaborative structures in the field of AI for wastewater treatment, an analysis of institutional productivity and co-authorship networks was conducted.

3.4.1 Leading Research Institutions

The productivity of research institutions was quantified by their total number of publications (TP) within the collected dataset. Table 4 lists the top 10 most productive institutions in this field.

Table 4

Top 10 Most Productive Research Institutions in AI for Wastewater Treatment Research (1987–2024)
Rank	Institutions	TP	Country
1	King Fahd University of Petroleum and Minerals	73	Saudi Arabia
2	Tsinghua University	51	China
3	Tongji University	48	China
4	King Khalid University	43	Saudi Arabia
5	Duy Tan University	40	Vietnam
6	Central South University	38	China
7	University Tehran	38	Iran
8	Prince Sattam Bin Abdulaziz University	36	Saudi Arabia
9	South China University of Technology	34	Chine
10	King Abdulaziz University	33	Saudi Arabia

As detailed in Table 4, King Fahd University of Petroleum and Minerals (Saudi Arabia) stands out as the most prolific institution, contributing 73 publications. Chinese universities also demonstrate significant output, with Tsinghua University (TP = 51) and Tongji University (TP = 48) ranking second and third, respectively. The strong presence of Saudi Arabian institutions is further evidenced by King Khalid University (TP = 43), Prince Sattam Bin Abdulaziz University (TP = 36), and King Abdulaziz University (TP = 33) all featuring in the top 10. This underscores a considerable concentration of research activity in this field within both Saudi Arabia and China, each nation being home to four of the top 10 institutions.

Duy Tan University from Vietnam holds the fifth position with 40 publications, highlighting its notable research involvement. The University of Tehran (Iran), with 38 publications, also shows substantial productivity. The overall distribution indicates that a select group of institutions leads research output, significantly shaping the knowledge landscape in this domain.

3.4.2 Institutional Collaboration Network

To explore the collaborative ties between institutions, a co-authorship network was constructed using Bibliometrix, with network parameters detailed as follows: the visualization displays the top 50 institutions, applying an automatic layout, association normalization for link strength, and the Walktrap algorithm for community detection (clustering). Isolated nodes were removed, and a minimum of one edge was required for inclusion. The resulting network is presented in Fig. 7.

The institutional collaboration network Fig. 7 reveals distinct patterns of research partnerships. Several prominent clusters are evident, often centered on highly productive institutions. For instance, a significant purple-colored cluster features King Khalid University (Saudi Arabia) as a central node, indicating strong collaborative links with other Saudi Arabian institutions and select international partners. Similarly, an orange-colored cluster highlights the dense collaborations among Chinese institutions, with Tsinghua University and the University of Chinese Academy of Sciences appearing as key players; Duy Tan University (Vietnam) is notably integrated into this cluster, suggesting its research output is significantly bolstered by these international partnerships.

King Fahd University of Petroleum and Minerals (Saudi Arabia), the institution identified as the leader regarding volume of publications, was the anchor institution for another distinct (green-colored) collaboration group. Overall, the network indicates that leading research institutions also serve as hubs within broader collaborative networks that may be national and/or international in scope. Furthermore, the prominence of the institutions Tsinghua University, King Khalid University, and King Fahd University of Petroleum and Minerals within the network aligns with high publication volume, again suggestive that high research productivity typically aligns with expansive collaborative engagement.

3.5 Most Relevant Authors

Following the analysis of journals, countries, and institutions, this section focuses on the individual researchers who are key contributors to the field of AI in wastewater treatment. The analysis identifies the most productive and impactful authors, examines their publication trajectory over time, and maps their collaborative networks.

3.5.1 Leading Authors by Productivity and Impact

Writing‍‌‍‍‌ quality of authors has been measured by their overall publications (TP), total citations (TC) within the dataset, average citations per publication (TC/TP), and h-index. Table 5 presents the top 10 most influential authors ranking for these ‍‌‍‍‌parameters.

Table 5

Top 10 Most Relevant Authors in AI for Wastewater Treatment Research (1987–2024)
Rank	Author	Institution	TP	TC	TC/TP	h-index
1	WANG Y	HUANG JINGANG	51	736	14.43	15
2	LI J	YANG JIAQIAN	40	1419	35.48	17
3	WANG J	WANG JINGTING	39	258	6.62	8
4	LI Y	FU SONGZHE	38	723	19.03	14
5	ZHANG Y	WANG JINGTING	38	646	17	15
6	LIU Y	MA SHUAIYIN	37	696	18.81	12
7	WANG X	ZHANG KAI	36	1085	30.14	14
8	WANG Z	YUAN SHIDENG	35	473	13.51	12
9	LI X	LI XINGYANG	34	672	19.76	13
10	WANG H	HUANG JINGANG	31	414	13.35	12

As presented in Table 5, Wang Y emerges as the most productive author with 51 publications. Following closely are Li J with 40 publications and Wang J with 39 publications. While productivity is a key indicator, citation metrics reveal further nuances regarding impact. Notably, Li J, despite being second in TP, exhibits the highest total citations (TC = 1419), the highest average citations per publication (TC/TP = 35.48), and the highest h-index (h = 17) among the top 10 authors, underscoring the significant influence of this author's work. Similarly, Wang X also demonstrates considerable impact with a high TC/TP of 30.14 and an h-index of 14. Conversely, some highly productive authors may have a more moderate citation impact per paper, illustrating the diverse contribution profiles within the leading research cohort. The prevalence of common surnames such as Wang, Li, and Zhang among the top authors further highlights the importance of considering potential name disambiguation challenges.

3.5.2 Temporal Production Trends of Top Authors

To understand the research life cycle and publication dynamics of top authors, their yearly scientific output is shown in Fig. 8. The bubble sizes indicate the number of articles the author published in that year, while the color brightness of the bubbles represents the total citations received (TC per Year) by the articles published by that author in that year.

Figure 8 reveals diverse publication trajectories among the top authors. For instance, Li X appears as an early contributor to the field, with publications dating back to the mid-1990s, followed by more consistent output in recent years. However, the majority of the top 10 authors, including Wang Y, Li J, and Wang J, began publishing significantly in this domain more recently, typically from around 2014–2018 onwards, yet have achieved high productivity in a relatively short period. This aligns with the overall accelerated growth of the research field itself. The sustained or increasing publication output from these authors in recent years highlights their ongoing active engagement.

3.5.3 Author Collaboration Network

The collaborative relationships among authors were analyzed to map the social structure of research in this field. Figure 9 presents the co-authorship network, displaying the top 50 authors based on their collaborative activities. The network was generated using Bibliometrix, employing an automatic layout, association normalization for link strength, and the Walktrap algorithm for community detection. Node size is proportional to the author's total publications (TP), and link thickness represents the frequency of co-authorship.

The co-authorship network, Fig. 9 illustrates a landscape characterized by a large, densely connected main component, alongside a few smaller, detached groups. Many of the top-ranking authors from Table 5, such as Wang Y and Li J, are prominently positioned within this main component, often appearing as central nodes within distinct colored clusters. This indicates that these highly productive individuals are also deeply embedded in collaborative research groups. For example the red cluster appears to connect authors like Li J and Wang X, while the blue cluster is centered around Wang Y and Zhang Y. The presence of several distinct clusters within the main component suggests the existence of multiple active research teams or communities. The smaller, isolated groups (e.g., Comas J and Poch M; Cho K) represent authors or small teams working with fewer connections to the broader network shown. Overall, the network underscores a high degree of collaboration among the leading researchers in the field of AI for wastewater treatment.

3.6 Most Cited Documents

To identify the most impactful research contributions at the intersection of AI and wastewater treatment, the top 10 most globally cited documents within the dataset were analyzed. These publications often represent significant advancements, foundational reviews, or widely adopted methodologies that have substantially influenced the direction and development of the field. Table 6 lists these highly cited documents along with their authors, publication year, source journal, and total citation count (TC) within the current dataset.

Table 6

Top 10 Most Cited Documents on AI in Wastewater Treatment Research (1987–2024)
Rank	Title	Authors	Year	Source	TC
1	STATE OF THE ART FOR GENETIC ALGORITHMS AND BEYOND IN WATER RESOURCES PLANNING AND MANAGEMENT	NICKLOW J, REED P, SAVIC D, DESSALEGNE T, HARRELL L, CHAN-HILTON A, KARAMOUZ M, MINSKER B, OSTFELD A, SINGH A, ZECHMAN E	2010	JOURNAL OF WATER RESOURCES PLANNING AND MANAGEMENT	532
2	DEEPARG: A DEEP LEARNING APPROACH FOR PREDICTING ANTIBIOTIC RESISTANCE GENES FROM METAGENOMIC DATA	ARANGO-ARGOTY G, GARNER E, PRUDENT A, HEATH, LENWOOD S L, VIKESLAND P, ZHANG L	2018	MICROBIOME	457
3	ACTIVATED SLUDGE WASTEWATER TREATMENT PLANT MODELLING AND SIMULATION: STATE OF THE ART	GERNAEY K, VAN L M, HENZE M, LIND M, JORGENSEN S	2004	ENVIRONMENTAL MODELLING \& SOFTWARE	394
4	BACTERIAL COMMUNITY STRUCTURES ARE UNIQUE AND RESILIENT IN FULL-SCALE BIOENERGY SYSTEMS	WERNER J, KNIGHTS D, GARCIA M, SCALFONE, NICHOLAS B N, SMITH S, YARASHESKI K, CUMMINGS, THERESA A T, BEERS A, KNIGHT R, ANGENENT L	2011	PROCEEDINGS OF THE NATIONAL ACADEMY OF SCIENCES OF THE UNITED STATES OF AMERICA	375
5	THE APPLICATION OF MACHINE LEARNING METHODS FOR PREDICTION OF METAL SORPTION ONTO BIOCHARS	ZHU X, WANG X, OK Y	2019	JOURNAL OF HAZARDOUS MATERIALS	244
6	AN INSIGHT INTO MACHINE LEARNING MODELS ERA IN SIMULATING SOIL, WATER BODIES AND ADSORPTION HEAVY METALS: REVIEW, CHALLENGES AND SOLUTIONS	YASEEN Z	2021	CHEMOSPHERE	239
7	AUTOMATED DETECTION OF SEWER PIPE DEFECTS IN CLOSED-CIRCUIT TELEVISION IMAGES USING DEEP LEARNING TECHNIQUES	CHENG J, WANG M	2018	AUTOMATION IN CONSTRUCTION	232
8	A DEEP LEARNING CNN ARCHITECTURE APPLIED IN SMART NEAR-INFRARED ANALYSIS OF WATER POLLUTION FOR AGRICULTURAL IRRIGATION RESOURCES	CHEN H, CHEN A, XU L, XIE H, QIAO H, LIN Q, CAI K	2020	AGRICULTURAL WATER MANAGEMENT	228
9	APPLICATION OF THE ANALYTIC HIERARCHY PROCESS AND THE ANALYTIC NETWORK PROCESS FOR THE ASSESSMENT OF DIFFERENT WASTEWATER TREATMENT SYSTEMS	BOTTERO M, COMINO E, RIGGIO V	2011	ENVIRONMENTAL MODELLING AND SOFTWARE	199
10	FUEL PROPERTIES OF HYDROCHAR AND PYROCHAR: PREDICTION AND EXPLORATION WITH MACHINE LEARNING	LI J, PAN L, SUVARNA M, TONG Y, WANG, XIAONAN X	2020	APPLIED ENERGY	195

The analysis of the most cited documents, as presented in Table 6, reveals a diverse set of influential works. The leading publication by Nicklow et al. (2010) [31], titled "State of the art for genetic algorithms and beyond in water resources planning and management," has garnered 532 citations, highlighting the enduring relevance of optimization techniques, including early AI approaches like genetic algorithms, in the broader water resources sector. Another highly cited earlier work is by Gernaey et al. (2004) [32], on activated sludge wastewater treatment plant modeling, which serves as a foundational piece for subsequent AI applications in process control and optimization.

A significant number of the top-cited articles are more recent, particularly those published from 2018 onwards, reflecting the rapid advancements and growing interest in contemporary AI techniques. For instance, Arango-Argoty et al. (2018) [33], on the application of deep learning for predicting antibiotic resistance genes (TC = 457), and also (Yaseen Z., 2021) [34] with (TC = 239) and Cheng & Wang (2018) on using deep learning for automated sewer pipe defect detection (TC = 232) showcase the impact of sophisticated AI models in addressing specific wastewater-related challenges. Similarly, publications focusing on machine learning applications, such as Bottero et al., (2011) [35] with (TC = 199) and Li et al. (2020) [25] (TC = 195)for predicting fuel properties of hydrochar/pyrochar, also feature prominently.

The thematic scope of these highly cited papers is broad, encompassing AI applications in water resources planning, prediction of emerging contaminants like antibiotic resistance genes, wastewater process modeling, analysis of microbial communities (Werner et al., 2011) [36], contaminant removal, infrastructure assessment, and resource recovery from waste. This diversity is also mirrored in the publication outlets, with influential articles appearing in specialized journals across water resources management (Journal of Water Resources Planning and Management), microbiology (Microbiome), environmental modeling (Environmental Modelling & Software), general high-impact science (PNAS), hazardous materials (Journal of Hazardous Materials), environmental chemistry (Chemosphere), automation (Automation in Construction), agricultural water management (Agricultural Water Management), and energy (Applied Energy).

3.7 Analysis of Keywords

A systematic examination of author keywords presents the general picture of the research domain's central themes and intellectual foundation. Such an examination reveals its leading research themes and time trends.

3.7.1 Most Frequent Keywords

A preliminary evaluation was conducted to determine the most commonly utilized author keywords, and its results are represented in Fig. 10. The term "machine learning" was identified as the leading term, utilized 562 times, consequently redefining its status as the leading methodology within this research area. Next, the important subject "wastewater treatment" was referred 185 times, followed by the generic concept of "artificial intelligence," which was utilized 183 times. More frequently cited keywords are advanced methods like "deep learning" referred 144 times, and "artificial neural network" referred 87 times, in addition to key operational terms like "optimization," referred 84 times, and "prediction," which was referred 64 times.

To illustrate the value and hierarchical structure of these topics, a treemap was generated from the co-occurrence frequencies of keyword bigrams from the abstracts of research articles (as shown in Fig. 11). The treemap highlights significantly the prevalence of "machine learning," which occupies the largest proportional area, sharing 21% of all areas. Next in ranking are the application areas of "wastewater treatment" (7%) and the novel method "deep learning" (5%). The treemap accurately distinguishes the major sub-topic constituents embedded within each broader category; for example, the "wastewater" category contains particular applications and media like "adsorption" and "water quality," while the AI category contains different methodologies like "artificial neural network" and "prediction," thus revealing the subtle hierarchy of valuable research emphases in the area.

3.7.2 Keyword Co-occurrence Network

To examine the relationships and intellectual associations between different concepts, a keyword co-occurrence network was created, as shown in Fig. 12. Keywords are represented as nodes in this network, and the edges between them indicate their simultaneous occurrences in the literature, revealing the structural basis of the research. Network analysis produced three distinct, strongly connected clusters.

The first cluster, indicated in green, refers to "machine learning" and its collection of algorithms and models, like "support vector," "random forest," and "decision tree," and thus reveals its methodological nature. The second cluster, colored in red, corresponds to "wastewater treatment" and related concepts like "activated sludge," "water quality," and "oxygen demand," and thus reveals its area of interest. The third cluster, marked in blue, concerns "neural network" and "artificial intelligence," which are important connectors between methodology-centered and application-based disparate clusters. The high number of interlinks between central nodes from all three clusters, that is, "machine learning," "wastewater treatment," and "neural network," reveal this research area's fundamentally interdisciplinary nature.

3.8 Thematic Map

Thematic mapping was employed to position research themes based on two criteria, centrality (relevance to the whole field) and density (how developed the theme is in the field) as shown in Fig. 13 below. The four quadrants of this thematic map visually divide the research landscape and provide an overview of the status and role of research themes within that landscape.

Motor Themes (Upper-Right Quadrant): These include elements that are highly developed and were identified as important to the research field such as "neural network", "artificial intelligence", and "support vector". These developments point to established and important themes necessary for the advancement of the discipline.
Basic Themes (Lower-Right Quadrant): This is where core aspects for the research field exist that have importance but plenty of room for growth. The core themes of "machine learning", "wastewater treatment", and "treatment plant" are located here, demonstrating high centrality and a place for future research.
Niche Themes (Upper-Left Quadrant): These are highly developed but specialized elements with less impact on the field. Things like "artificial neural network ann" and "removal efficiency" appeared here, locating them as mature subjects of specific research streams.
Emerging or Declining Themes (Lower-Left Quadrant): Themes located in this quadrant have low development and centrality. Things like "random forest", "gradient boosting", and "decision support system" are located here. These could either signify emerging niche methods that have not yet become more centrally located or they indicate older paradigms that are being replaced by the more comprehensive "machine learning" paradigm.

3.9 Thematic Evolution

In order to evaluate the shifts in thematic emphasis throughout time, a thematic change analysis was performed to consider the periods of 1987–2015 and 2016–2024. Figure 14 depicts a Sankey diagram to illustrate the changes in research themes from one period to the next.

In the first period (1987–2015), the first authors focused their research on fundamental themes like "wastewater treatment," "artificial neural network," "optimization," and "decision support systems." These inventions formed a foundation on the initial work undertaken using computational intelligence to further wastewater treatment research projects.

The second period (2016–2024) indicates that research themes have been significantly redeveloped and consolidated. The flow chart indicates that many of those previous themes, such as "artificial neural networks" and "optimization," have now been encompassed under the broader and growing theme of "machine learning." What this conveys is an advancement in what is now expected as the encompassing framework. In addition, it is noteworthy that there remains the emergence of "deep learning" as a significant independent theme, which has evolved from "artificial neural networks." Further, it is evident that this evolutionary chart exhibited a scientific trajectory from discrete models of computational intelligence toward encompassing platforms of machine learning, to a rise and evolution of complex models of deep learning.

This bibliometric analysis of 2,096 journal articles published in Scopus and Web of Science from 1987 to 2024 shows how the research community studying the use of Artificial Intelligence (AI) for wastewater treatment has progressed in recent decades. The findings show that AI has experienced exponential growth (average growth is 17% per year), which is further evidenced by a strong collaborative profile (mean of 5.64 co-authors per article; 25.1% of articles had an international co-author), and a relatively young body of research (average age of 4 years) that is demonstrating an average citation rate of 17.63 citations per article. These trends not only indicate who the key players are and that thematic shifts have occurred over time, but also indicate that AI is no longer perceived as a peripheral approach for solving wastewater treatment problems, but rather is an established methodological pillar for solving wastewater solutions.

4.1 Growth Dynamics and the Emergence of a Mature Research Field

Our findings reveal a nearly exponential increase in publications beginning in 2016 and an annual increase of 17.14%. Like several other recent bibliometric studies, we found evidence to strongly support this finding. Yu et al., [37], for instance, noted a "sharply" increasing number of papers published after 2018 and Li et al., [38] noted rapid development after 2020 and described it as "an explosive growth trend." This acceleration corresponds with major developments in deep learning (to inform prediction), the availability of big data from modern wastewater treatment plants, and accessibility to compute power [39]. The slow build up of the field prior to 2008, the steady growth that followed, and then explosive growth is also a textbook example of scientific growth.

4.2 Collaboration Landscape and Scientific Leadership

The examined geography suggests a preference for dominant production countries in terms of research output, namely China, USA, and India. Studies from Yu et al. (2023) [37], Baarimah et al. (2024) [40] and Li et al. (2024) [38] support the order of countries produced prolific research output. Our findings provide some additional detail on this finding. For instance, our study identified Spain as the highest average citations per paper (10.47). While they do not have the same number of researchers, they produce quality research output, which is in line with the previous research on water resources research which identified several European countries as producing high quality research as well [41, 14].

In the same, Li et al. (2024) [38] indicates that the number of publications from China is high, but the average citation per publication from China is relatively low. This implies that the more meaningfully impactful research that has the ability to close the gap between quantity and quality is occurring in other locations. This demonstrates that there are a variety of different research strategies present in different countries: China relies on its large domestic research network; countries like Saudi Arabia rely on international networks; and the USA is clearly a central location in the global network [15, 19, 29]. Closer to home, emerging collaborative efforts in Southeast Asia and the Middle East demonstrate a gradual diversification of global applications of AI on wastewater management, as each of these developing countries work more frequently on collaborative research with other institutions that have distinguished track records in this space [5, 21, 13, 18]. This emerging connectedness constitutes a more mature global ecosystem, with an increasing degree of exchange of expertise, data, and methodologies resulting in an evolution of intelligent and sustainable wastewater management systems worldwide.

4.3 Thematic Evolution: From Models to Methods

Perhaps one of the key indicators from our study is the evolution of keywords. We move from a model-specific paradigm, such as "artificial neural networks" (ANN), to an umbrella framework (machine learning) connected by the complexity of wastewater problem x, and especially emerging contaminants. Emerging contaminants (ECs), like pharmaceuticals and antibiotics, are an emerging research area [10] that has driven methodological evolution to improve outcomes. Research hotspots have evolved instead of just AI techniques like "support vector machine" and "random forests," include specific areas of application like "membrane bioreactors" and "photocatalytic degradation" where AI models are implemented to optimize outcomes [38]. Procedures implemented through modern deep learning applications are solving the more complex problems of predicting antibiotic resistance genes [33] or considering the automatic inspection of infrastructures like dams [8]. The deep learning field has evolved into a connection of solving real complex environmental problems rather than theory by exploring many AI models.

4.4 Dissemination Channels and Research Impact

The prominence of this work in Science of the Total Environment, Journal of Environmental Management, and Water Research indicates that AI-enabled treatment of wastewater-related research is increasingly published in high-impact outlets. For instance, Zhou et al. (2025) [42] showed that nine out of ten of these journals are Q1 by the journal metrics for their respective disciplines. Topic (or discipline) significance is also evident by Li et al. (2024) [38] inclusion of the same three journals among productive journals, illustrating where research dissemination is being organized for AI-environmental research. Similarly, it is worth highlighting the prominence of multidisciplinary environmental journals that facilitate cross-pollination between applied computational intelligence and environmental engineering Ray [25, 43, 4].

The relevance of foundational review papers, that still receive high citation counts, emphasizes the continuing significance of optimization and modeling approaches to these challenges in (6, 27) (20). In parallel, the recent rise in citation counts signals a rapidly accumulating body of literature on applied AI in the resource recovery context of wastewater treatment, enabled by emerging hybrid models and deep learning systems [17, 44, 10, 45]. The distribution of research across journals like Water Science and Technology, Journal of Water Process Engineering, and Environmental Science & Technology [3, 26, 46] suggests research outputs form an connected knowledge base, while simultaneously employing a diverse lens. Together, these research outputs will facilitate a global research impact that is multifaceted and interdisciplinary, and should enable AI-led wastewater management to truly be recognized as a centerpiece of sustainable innovation for environmental science.

4.5 Implications and Future Perspectives

This analysis and the findings of other research provide clear evidence that the use of AI in wastewater treatment is a vibrant, rapidly growing research area. The technological impetus for the growth in AI application is situated within the broader context of a global shift towards a circular economy, where wastewater is increasingly viewed as a resource (for water, energy, and nutrients), not as a waste product [47]. In this context, AI is positioned as an important enabler for the transformation of treatment facilities into resource recovery facilities. Future research will likely pursue some important areas. There is evidence of a movement towards greater applications of sophisticated "hybrid algorithms" that complement existing AI models and mechanistic models, with a goal of improving predictions or control of systems [38].

Secondly, an important future direction is using machine learning to address emerging contaminants. Chen et al., [10] highlight one promising idea is using ML to identify unknown ECs in wastewater, going beyond treating known pollutants. Thirdly, AI is evolving from process optimization to material science, with a key future direction being the AI-enabled design of new materials, such as new adsorbents or catalysts for wastewater treatment [38]. Finally, to move the field forward research needs to extend beyond laboratory scale data and start to use real-world data from operational wastewater treatment plants to produce more robust and holistic algorithms that consider the full treatment process [38].

A comprehensive bibliometric evaluation of nearly forty years of research on the use of Artificial Intelligence in wastewater treatment demonstrates the field is dynamic and rapidly maturing, has tremendous growth, increased co-authorship globally, and a specific move away from modeling practices for hybrid, data-driven, and systems-based approaches. A synthesis of 2,096 journal articles in Scopus and Web of Science spanning from 1987 to 2024 presents evidence showing the evolution AI research has taken, from being an outlier practice involving the creation of one or two models, to a vital epistemological position, epistemic commitment, and accountability for sustainable wastewater management systems. This evolution is reported not only as output in science and annual growth rate (17.14%) of published literature after 2016, but also as a more significant diversity of themes, techniques, and applications that together build the intellectual architecture of the field. The data presented build/support the growing evidence that AI is moving away for being just a niche analytical technique for modeled systems into a mainstream pathway for enablers of sustainability-based wastewater management.

The world map of research in this field reveals considerable asymmetries in research production and knowledge production. China is the leading producer, while countries such as Spain, Iran, and South Korea have a disproportionately high citation impact given their publication volume and signify that research excellence and visibility do not strictly depend on quantity of output [38, 42]. These findings are consistent with wider scientometric patterns pertaining to the environmental sciences and water science and suggest that, for example, European and Middle Eastern countries are not producing much, but when they do, their studies are cited more often [48]. At the same time, heavy producer countries such as China and India are developing strong domestic networks to build a base of publications and position themselves to be leaders in technology. The range of collaboration strategies utilized, from China and Spain's focused domestic research to Saudi Arabia and Australia's highly internationalized networks, highlights that development of scientific leadership in AI research in wastewater treatment is dependent on both research capacity and connectivity across global knowledge systems.

In terms of themes, this examination indicates a significant conceptual change in the literature, as previous works focused on distinct computable models such as artificial neural networks, decision support systems, and genetic algorithms as they relate to process optimization and parameterization [31, 48]. In time, these disparate methodological approaches have come together within machine learning, including supervised and unsupervised learning methods, ensemble approaches, and deep learning frameworks. This transition is more than a change of terminology it is the advancement of the arena of wastewater research into a data- and prediction-driven field. Deep learning models have now been used successfully in ways that were once thought impossible like automatically classifying sewer defects from image recognition [8] or surfacing antibiotic resistance targets from metagenomic data [33].These novel developments have demonstrated that AI tools are not just improving existing tasks but, are actually transforming the limits of what a wastewater treatment systems can accomplish.

These findings have far-reaching implications that extend to the more general topic of sustainability. Wastewater is increasingly understood as a recoverable resource through the circular economy model of recovering, re-using, and repurposing simultaneously energy, water and nutrients [47]. In this context, AI represents a paradigm shift opportunity for systems integration, process optimization, real-time monitoring and adaptive control. Machine learning models can enhance the accuracy of effluent quality projections, optimize aeration practices to improve energy efficiencies, and support predictive maintenance practices to reduce operational costs. Further, the combination of AI with big data analytics and the Internet of Things [39]. AI could radically advance the vision of decentralized, smart wastewater management systems. These advances directly contribute to achieving the United Nations Sustainable Development Goals (SDGs) related to water and sanitation (SDG 6) and responsible production and consumption (SDG 12), although are increasingly seen (along with the SDGs) as providing efficiency, as well as resilience and environmental protection.

At the same time, the findings reveal structural and epistemic gaps that offer potential for future research. A significant gap involves limited, large-scale validation of AI models in the real world. That means, partnerships, studies, and real-world data focused on AI models, currently tend to either rely on experimentation in the lab or "ideal" or "unique" datasets that may not accurately resemble heterogeneity, uncertainty and noise when the system is performed at full-scale [38]. To fill this gap, collaboration among researchers, industry practitioners, and utilities would begin to create shared data infrastructures that could support validation of AI models for the purpose of detailed model training and evaluation. Another important frontier of research is in creating explainable AI (XAI) to build understanding of model outputs, and trust, by wastewater plant operators. Most applications of AI technologies are based on opaque "black box" models. These "black box" models are accurate, but inhibit transparency, often preventing enhanced user engagement. XAI offers an opportunity to develop human and machine collaboration and to ensure AI recommendations reflect process knowledge and safety factors.

Equally significant is the examination of hybrid modeling approaches that combine AI with mechanistic process models. These hybrid systems can leverage machine learning's predictive ability sufficient while retaining the physical interpretability of traditional process models, allowing for both accuracy and interpretability [38]. Moreover, the use of AI in wastewater treatment has mainly focused on optimization of pollutant removal, while optimizing for holistic resource recovery such as nab(not including, or segregating) optimization of nutrient recovery, energy recovery, and water reuse has not received significant attention [10]. Addressing these research gaps could hasten transitioning wastewater treatment plants to multi-products or multi-functional resource recovery facilities that uphold circular economy and sustainable resource management principles.

From a policy and governance viewpoint, the bibliometric evidence provided in this article provides useful knowledge for decision-makers. The fact that the research activity sits within a few dominant countries suggest the necessity for greater emphasis on collaborative, international, and technology transfer efforts to ensure equitable entry to AI developments. Funding agencies can use this information to promote funding in regions of underrepresentation and emerging lines of inquiry, like explainable AI, scale in the real-world, and hybrid modeling. Journal editors as well as researchers can use the identified patterns of citation and collaboration, to strengthen dialogue across fields and disciplines. They may additionally recognize the benefit of promoting the opening of data to create greater transparence and reproducibility of findings.

In summary, Artificial Intelligence has gone from an experimental approach to a mature, foundational system for wastewater treatment research and practice. Integrating AI and the Environmental Engineering field represents a remarkable transformation toward intelligent, adaptive, and sustainable systems. The rapid growth of the field, tremendous international collaboration, and increased thematic concentration are strong indicators that innovative AI-based approaches will not only serve as additional computational tools but also help re-conceptualize wastewater as a recoverable and valuable resource. Addressing the compounded challenges of climate change, water scarcity, and pollution, responsible progress in the advancement and application of AI technologies will be important for accomplishing a resilient, efficient, and circular water future. Overall, the implications of this bibliometric review provide both a record of scientific progress and a pathway for future research and innovation at the intersection of artificial intelligence and wastewater management.

Turkdogan-Aydınol FI, Yetilmezsoy K. A fuzzy-logic-based model to predict biogas and methane production rates in a pilot-scale mesophilic UASB reactor treating molasses wastewater. J Hazard Mater. oct 2010;182(1‑3):460‑71.
Akratos CS, Papaspyros JNE, Tsihrintzis VA. An artificial neural network model and design equations for BOD and COD removal prediction in horizontal subsurface flow constructed wetlands. Chem Eng J. sept 2008;143(1‑3):96‑110.
Sigmund G, Gharasoo M, Hüffer T, Hofmann T. Deep Learning Neural Network Approach for Predicting the Sorption of Ionizable and Polar Organic Pollutants to a Wide Range of Carbonaceous Materials. Environ Sci Technol. 7 avr 2020;54(7):4583‑91.
Wang D, Thunéll S, Lindberg U, Jiang L, Trygg J, Tysklind M, et al. A machine learning framework to improve effluent quality control in wastewater treatment plants. Sci Total Environ. août 2021;784:147138.
Zhu X, He M, Sun Y, Xu Z, Wan Z, Hou D, et al. Insights into the adsorption of pharmaceuticals and personal care products (PPCPs) on biochar and activated carbon with the aid of machine learning. J Hazard Mater. févr 2022;423:127060.
Jia H, Yao H, Tang Y, Yu SL, Field R, Tafuri AN. LID-BMPs planning for urban runoff control and the case study in China. J Environ Manage. févr 2015;149:65‑76.
Nourani V, Elkiran G, Abba SI. Wastewater treatment plant performance analysis using artificial intelligence – an ensemble approach. Water Sci Technol. 21 déc 2018;78(10):2064‑76.
Cheng JCP, Wang M. Automated detection of sewer pipe defects in closed-circuit television images using deep learning techniques. Autom Constr. nov 2018;95:155‑71.
Sun X, Wang P, Wang C, Liu Y, Fu K. PBNet: Part-based convolutional neural network for complex composite object detection in remote sensing imagery. ISPRS J Photogramm Remote Sens. mars 2021;173:50‑65.
Chen Y, Lin M, Zhuang D. Wastewater treatment and emerging contaminants: Bibliometric analysis. Chemosphere. juin 2022;297:133932.
Li J, Pan L, Suvarna M, Wang X. Machine learning aided supercritical water gasification for H2-rich syngas production with process optimization and catalyst screening. Chem Eng J. déc 2021;426:131285.
Frombo F, Minciardi R, Robba M, Sacile R. A decision support system for planning biomass-based energy production. Energy. mars 2009;34(3):362‑9.
Li J, Zhu X, Li Y, Tong YW, Ok YS, Wang X. Multi-task prediction and optimization of hydrochar properties from high-moisture municipal solid waste: Application of machine learning on waste-to-resource. J Clean Prod. janv 2021;278:123928.
Bagherzadeh F, Mehrani MJ, Basirifard M, Roostaei J. Comparative study on total nitrogen prediction in wastewater treatment plant and effect of various feature selection methods on machine learning algorithms performance. J Water Process Eng. juin 2021;41:102033.
Khan M, Raza Naqvi S, Ullah Z, Ali Ammar Taqvi S, Nouman Aslam Khan M, Farooq W, et al. Applications of machine learning in thermochemical conversion of biomass-A review. Fuel. janv 2023;332:126055.
Zhang J, Liu J, Evrendilek F, Zhang X, Buyukada M. TG-FTIR and Py-GC/MS analyses of pyrolysis behaviors and products of cattle manure in CO2 and N2 atmospheres: Kinetic, thermodynamic, and machine-learning models. Energy Convers Manag. sept 2019;195:346‑59.
Zhu X, Wan Z, Tsang DCW, He M, Hou D, Su Z, et al. Machine learning for the selection of carbon-based materials for tetracycline and sulfamethoxazole adsorption. Chem Eng J. févr 2021;406:126782.
Wang G, Qiao J, Bi J, Li W, Zhou M. TL-GDBN: Growing Deep Belief Network With Transfer Learning. IEEE Trans Autom Sci Eng. avr 2019;16(2):874‑85.
Liu X, Shao Z, Wang Y, Liu Y, Wang S, Gao F, et al. New use for Lentinus edodes bran biochar for tetracycline removal. Environ Res. janv 2023;216:114651.
Molinos-Senante M, Garrido-Baserba M, Reif R, Hernández-Sancho F, Poch M. Assessment of wastewater treatment plant design for small communities: Environmental and economic aspects. Sci Total Environ. juin 2012;427‑428:11‑8.
Granata F, Papirio S, Esposito G, Gargano R, De Marinis G. Machine Learning Algorithms for the Forecasting of Wastewater Quality Indicators. Water. 9 févr 2017;9(2):105.
Manhou K, Moussadek R, Yachou H, Zouahri A, Douaik A, Hilal I, et al. Assessing the Impact of Saline Irrigation Water on Durum Wheat (cv. Faraj) Grown on Sandy and Clay Soils. Agronomy. 1 déc 2024;14(12):2865.
Manhou K, Taghouti M, Moussadek R, Elyacoubi H, Bennani S, Zouahri A, et al. Performance, Agro-Morphological, and Quality Traits of Durum Wheat (Triticum turgidum L. ssp. durum Desf.) Germplasm: A Case Study in Jemâa Shaïm, Morocco. Plants. 17 mai 2025;14(10):1508.
Manhou K, Moussadek R, Dakak H, Zouahri A, Ghanimi A, Sanad H, et al. Comparative assessment of saline irrigation effects on germination, physiological responses, growth, and yield of durum wheat varieties grown on silty clay soil [Internet]. In Review; 2025 [cité 30 oct 2025]. Disponible sur: https://www.researchsquare.com/article/rs-7909033/v1
Li J, Pan L, Suvarna M, Tong YW, Wang X. Fuel properties of hydrochar and pyrochar: Prediction and exploration with machine learning. Appl Energy. juill 2020;269:115166.
Wang G, Jia QS, Zhou M, Bi J, Qiao J, Abusorrah A. Artificial neural networks for water quality soft-sensing in wastewater treatment: a review. Artif Intell Rev. janv 2022;55(1):565‑87.
Poch M, Comas J, Rodríguez-Roda I, Sànchez-Marrè M, Cortés U. Designing and building real environmental decision support systems. Environ Model Softw. sept 2004;19(9):857‑73.
Dellana SA, West D. Predictive modeling for wastewater applications: Linear and nonlinear approaches. Environ Model Softw. janv 2009;24(1):96‑106.
Wang D, Thunéll S, Lindberg U, Jiang L, Trygg J, Tysklind M. Towards better process management in wastewater treatment plants: Process analytics based on SHAP values for tree-based machine learning methods. J Environ Manage. janv 2022;301:113941.
Wang F, Wang Y, Zhang K, Hu M, Weng Q, Zhang H. Spatial heterogeneity modeling of water quality based on random forest regression and model interpretation. Environ Res. nov 2021;202:111660.
Nicklow J, Reed P, Savic D, Dessalegne T, Harrell L, Chan-Hilton A, et al. State of the Art for Genetic Algorithms and Beyond in Water Resources Planning and Management. J Water Resour Plan Manag. juill 2010;136(4):412‑32.
Gernaey KV, Van Loosdrecht MCM, Henze M, Lind M, Jørgensen SB. Activated sludge wastewater treatment plant modelling and simulation: state of the art. Environ Model Softw. sept 2004;19(9):763‑83.
Arango-Argoty G, Garner E, Pruden A, Heath LS, Vikesland P, Zhang L. DeepARG: a deep learning approach for predicting antibiotic resistance genes from metagenomic data. Microbiome. déc 2018;6(1):23.
Yaseen ZM. An insight into machine learning models era in simulating soil, water bodies and adsorption heavy metals: Review, challenges and solutions. Chemosphere. août 2021;277:130126.
Bottero M, Comino E, Riggio V. Application of the Analytic Hierarchy Process and the Analytic Network Process for the assessment of different wastewater treatment systems. Environ Model Softw. oct 2011;26(10):1211‑24.
Werner JJ, Knights D, Garcia ML, Scalfone NB, Smith S, Yarasheski K, et al. Bacterial community structures are unique and resilient in full-scale bioenergy systems. Proc Natl Acad Sci. 8 mars 2011;108(10):4158‑63.
Yu X, Chen S, Zhang X, Wu H, Guo Y, Guan J. Research progress of the artificial intelligence application in wastewater treatment during 2012–2022: a bibliometric analysis. Water Sci Technol. 1 oct 2023;88(7):1750‑66.
Li X, Su J, Wang H, Boczkaj G, Mahlknecht J, Singh SV, et al. Bibliometric analysis of artificial intelligence in wastewater treatment: Current status, research progress, and future prospects. J Environ Chem Eng. août 2024;12(4):113152.
Aliouache M, Wang X, Fischer P, Massonnat G, Jourde H. An inverse approach integrating flowmeter and pumping test data for three-dimensional aquifer characterization. J Hydrol. déc 2021;603:126939.
Baarimah AO, Bazel MA, Alaloul WS, Alazaiza MYD, Al-Zghoul TM, Almuhaya B, et al. Artificial intelligence in wastewater treatment: Research trends and future perspectives through bibliometric analysis. Case Stud Chem Environ Eng. déc 2024;10:100926.
Pasqualino JC, Meneses M, Abella M, Castells F. LCA as a Decision Support Tool for the Environmental Improvement of the Operation of a Municipal Wastewater Treatment Plant. Environ Sci Technol. 1 mai 2009;43(9):3300‑7.
Zhou K, Wu B, Zhang X. Worldwide Research Progress and Trends in Application of Machine Learning to Wastewater Treatment: A Bibliometric Analysis. Water. 28 avr 2025;17(9):1314.
Sharafati A, Asadollah SBHS, Hosseinzadeh M. The potential of new ensemble machine learning models for effluent quality parameters prediction and related uncertainty. Process Saf Environ Prot. août 2020;140:68‑78.
Sarkar DJ, Das Sarkar S, Das BK, Sahoo BK, Das A, Nag SK, et al. Occurrence, fate and removal of microplastics as heavy metal vector in natural wastewater treatment wetland system. Water Res. mars 2021;192:116853.
Xie W, Li T, Tiraferri A, Drioli E, Figoli A, Crittenden JC, et al. Toward the Next Generation of Sustainable Membranes from Green Chemistry Principles. ACS Sustain Chem Eng. 11 janv 2021;9(1):50‑75.
Guo H, Jeong K, Lim J, Jo J, Kim YM, Park J pyo, et al. Prediction of effluent concentration in a wastewater treatment plant using machine learning models. J Environ Sci. juin 2015;32:90‑101.
Camón Luis E, Celma D. Circular Economy. A Review and Bibliometric Analysis. Sustainability. 7 août 2020;12(16):6381.
Lin H, Lin S, Awasthi MK, Wang Y, Xu P. Exploring the bacterial community and fermentation characteristics during silage fermentation of abandoned fresh tea leaves. Chemosphere. nov 2021;283:131234.
Hamed MM, Khalafallah MG, Hassanien EA. Prediction of wastewater treatment plant performance using artificial neural networks. Environ Model Softw. oct 2004;19(10):919‑28.

The authors declare no competing interests.